Scratch cards guide
Crystal Scratch: how far results swing from the average
The crystal in this page's address is a naming choice and refers to no real card. The topic is variance: the spread of outcomes around the expected value, which on a scratch card is far wider than the average suggests.
Average is not the whole story
The expected value tells you the centre of the outcomes. Variance tells you how scattered they are. For a single ticket, variance is the average of the squared distance between each prize and the expected prize. Its square root, the standard deviation (SD), is in the same units as the prize, so it is easier to read.
The example run below is hypothetical: 400,000 tickets at 4 units, with two top prizes of 40,000 units.
| Prize | Tickets | Odds | Expected value per ticket |
|---|---|---|---|
| 40,000.00 | 2 | 1 in 200,000.0 | 0.2000 |
| 1,000.00 | 30 | 1 in 13,333.3 | 0.0750 |
| 100.00 | 900 | 1 in 444.4 | 0.2250 |
| 20.00 | 7,000 | 1 in 57.1 | 0.3500 |
| 8.00 | 24,000 | 1 in 16.7 | 0.4800 |
| 4.00 | 64,000 | 1 in 6.2 | 0.6400 |
95,932 winning tickets out of 400,000 (overall 1 in 4.17). Prize fund 788,000.00 on sales of 1,600,000.00: return 49.25%, so a ticket at 4.00 is worth 1.9700 on average. Hypothetical print run.
The numbers for one ticket
The expected prize is 1.9700 units. The standard deviation of the prize is 90.04 units. That SD is many times larger than the price of the ticket, which tells you the outcome is dominated by rare, large payouts. A single ticket is almost always a loss or a small refund, and very rarely something huge.
To see which tier drives this, remove the two top prizes and recompute. The SD falls to 10.38 units. Two tickets out of 400,000 account for most of the spread, because variance squares each distance from the mean, and 40,000 units is a very long way from the mean.
Spread over many tickets
For n tickets bought independently, the expected net grows in proportion to n, but the SD grows only with the square root of n. That is why the average loss becomes clearer as purchases pile up.
| Tickets | Expected net | SD of net | SD without the top tier |
|---|---|---|---|
| 1 | -2.03 | 90.04 | 10.38 |
| 25 | -50.75 | 450.19 | 51.91 |
| 250 | -507.50 | 1,423.64 | 164.14 |
| 2,500 | -5,075.00 | 4,501.95 | 519.05 |
Treats tickets as independent draws, which is close enough for purchases far smaller than the run.
Why a normal-curve band misleads here
With dice or coin flips, a rough rule says most results land within two SDs of the average. That rule assumes a fairly symmetric spread. Scratch-card prizes are extremely lopsided: almost all of the variance sits in prizes that most buyers never see. For a few hundred tickets, the typical buyer ends up close to the expected loss computed from the lower tiers alone, and the large SD only reflects the tiny chance of a top prize.
A more honest description for one person is the median result, which here is a loss. Compare the last two columns of the table: the right-hand one is a better guide to the swings a real buyer feels.
Takeaways
- A large SD does not mean you are likely to win big. It means a few tickets are worth a great deal.
- Over more tickets, the average loss of 2.0300 units per card becomes the main feature.
- For streak and ruin calculations on house games, see the house edge guide.